Abstract
Critical raw material (CRM) exploration increasingly combines Earth Observation, geophysics, geochemistry, mineral prospectivity mapping, probabilistic modelling, and autonomous sensing. However, these capabilities are commonly developed or deployed as separate components, while the explicit coupling of observation, uncertainty-aware geological inference, and subsequent exploration decisions remains comparatively underdeveloped. This paper presents Closed-Loop Mineral Intelligence (CLMI) as a conceptual and methodological framework for organising these capabilities within an iterative Observe–Model–Decide process. The contribution does not propose a new stand-alone learning algorithm or claim an operationally validated exploration system. Instead, it formalises a reference architecture in which heterogeneous observations update probabilistic geological representations, uncertainty informs the prioritisation of subsequent observations, and expert oversight remains part of the decision loop. The framework is examined through an illustrative lithium-bearing pegmatite scenario that specifies the intended information flow, candidate data sources, modelling functions, decision points, and human intervention without reporting empirical performance. A validation roadmap is further defined to make the framework testable in future case studies, including requirements for benchmark baselines, uncertainty calibration, information-gain assessment, exploration-efficiency metrics, computational feasibility, and human–AI interaction. By distinguishing established technological components from the proposed system-level integration, the paper positions CLMI as a research framework and agenda for adaptive, uncertainty-aware mineral exploration rather than as a demonstrated replacement for current professional practice.